All,
  Srini identified that VNFs might require a Real-time Service Registration for 
VNFs.

  I wanted to let you know that AAF already has a real time Service Locator, 
that includes the following information:

   Name:  Hierarchical name of service – (i.e. org.onap.aai)
   hostname: FQDN
   port: (obvious)
   latitude: When more than one instance exists, use with longitude to 
determine closest service via Great Circle
   longitude:  When more than one instance exists, use with latitude to 
determine closest service via Great Circle
   version: (major, minor patch, pkg) to support increasing Versions realtime 
and pick best version of Service (i.e. A client that only knows about version 
1.0, will get a service with supports that Interface Version)
   protocol: ex: https
   subprotocol: extra info, example: “'TLSv1.1', 'TLSv1.2'”, indicating that 
client needs at least TLS 1.1 to talk to this service.

  How it might help you:

  1.  Current AAF version relies on Public FQDN, which, given K8s, would 
correlate to multiple K8s Service Implementations
  2.  Resilient and Global in scope (Lat/Long allows talking to remote service 
that is closest, when possible, but also allows contacting any service, if 
closer ones are not available)


  What is already happening in Dublin:

  1.  I am already designing/half implemented how to handle “K8S Internal 
names” (i.e. K8s <ns>.<service> i.e. onap.aaf_service) versus standard Public 
Names (FQNS known by DNS)

Please let me know if you interested in hearing about this existing element, 
what may work for you today, etc.


--
Jonathan Gathman
Principled-System Architect
ATO Tech Dev/SEAT/Platform Architecture and Technology Management

AT&T Services, Inc.
2349 Oaker, Arnold, MO 63010
m  314-550-3312  |  [email protected]<mailto:[email protected]>


From: <[email protected]> on behalf of Srini 
<[email protected]>
Date: Sunday, January 6, 2019 at 10:08 AM
To: "[email protected]" <[email protected]>, 
"[email protected]" <[email protected]>
Cc: "'de Talhouet, Alexis'" <[email protected]>, "MALAKOV, YURIY" 
<[email protected]>, "Ranganathan, Dileep" <[email protected]>, "Raj, 
Rajamohan" <[email protected]>, "Chivukula, Srivahni" 
<[email protected]>
Subject: [Onap-arc] Orchestration scenario : Inter-service parameters - 
Simplification question & seeking suggestions

Hi,

We have a requirement in one of the use cases we are working on.  I think that 
this scenario is generic and hence asking for community feedback.

Trying to understand whether this scenario can be realized with existing 
capabilities (such as CDS) or require some additional enhancements.

Scenario:

  *   An infrastructure network-service  (Infra) consisting of various VNFs and 
workloads in those VNFs.
  *   An app network-service (App) consisting of various VNFs and workloads in 
those VNFs.
  *   App is dependent on functionality offered by infra.
  *   Infra can be brought up in multiple cloud-regions.
  *   When App is brought up on a selected cloud-region, it is expected to use 
that cloud-region’s infra  workload functions via its API.
  *   For this discussion, let us assume infra and app network-services have 
only one VNF.

Challenge:

  *   When  user is instantiating App network-service, he/she needs to ensure 
that App VNF workloads are configured with right infra workload API endpoints, 
credentials and the stuff required to talk to infra workload webservices.
  *   Say that apps can be brought up on say any of 100 cloud-regions at later 
time, how does user know infra workload endpoint information for all these 
instances?

One out-of-band solution :

  *   User to maintain a document/spreadsheet with all endpoint information for 
various infra instances.
  *   User to refer to this document, while bringing up the app.

Problems with above approach:

  *   Manual and buggy.
  *   On technical front:  In some cases, at the time of app instantiation, 
user may not know the cloud-region and hence infra VNF instance on which app 
would be brought up on.  This case happens when dynamic cloud-region selection 
is taken care by OOF (using various constraints such as distance, cost, 
latency, HPA, affinity and other rules).

One possible solution (require enhancements in ONAP) is to maintain database of 
run time information of services on per VNF basis and leverage this when new 
service/VNF (app) is being instantiated). Something like this:


  1.  User brings up infra-service.
  2.  Since user knows the parameters (e.g API endpoints) this service exposes, 
let the user create the DB document with key as service+VNF+instance and 
document as set of parameters and values (via some GUI/CLI). Instance can be 
represented as “Cloud-region-id or customer id or combination of both”
  3.  User bring up app-service
  4.  For each VNF, ONAP will figure out the dependent service+VNF parameters 
(based on cloud-region selected by OOF) and use them to override values of 
app-service VNF parameters.

Even here, there are user actions that can lead to mistakes.  Automation of DB 
document creation is needed. For that, ONAP system need to know the parameters 
to put in. It means that we need to have some definition of parameters that are 
useful for other services that run on top of it.  Also, there should be some 
discipline on using consistent parameter names among VNFs of dependent services 
and let us assume that this discipline is assumed.

Does CDS help?

Or is this scenario be realized only after implementing nested services?

Thoughts and suggestions?

Thanks
Srini









-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.

View/Reply Online (#14814): https://lists.onap.org/g/onap-discuss/message/14814
Mute This Topic: https://lists.onap.org/mt/28962690/21656
Group Owner: [email protected]
Unsubscribe: https://lists.onap.org/g/onap-discuss/unsub  
[[email protected]]
-=-=-=-=-=-=-=-=-=-=-=-

Reply via email to